26.05 NeMo Retriever Library release notes#2065
Conversation
6a0f3a3 to
882713e
Compare
882713e to
647b30f
Compare
Greptile SummaryThis PR replaces the 26.03 body on
|
| Filename | Overview |
|---|---|
| docs/docs/extraction/releasenotes.md | 26.05 release notes added with subsection structure; one subsection heading missing before the text-to-SQL/tabular bullet, leaving it orphaned outside any named section. |
Flowchart
%%{init: {'theme': 'neutral'}}%%
flowchart TD
A[releasenotes.md] --> B["## 26.05 Release Notes (26.5.0)"]
B --> C["### Upgrade notes"]
B --> D["### Pipeline and ingestion"]
B --> E["### CLI"]
B --> F["### Retriever Service and deployment"]
B --> G["### Models, OCR, and captioning"]
B --> H["### Multimodal extraction"]
B --> I["### Retrieval and RAG"]
B --> J["### Vector database"]
B --> K["### Evaluation"]
B --> L["⚠️ ORPHANED BULLET — missing ### heading\n(Text-to-SQL and tabular)"]
B --> M["### Packaging and platform"]
B --> N["### Helm chart"]
B --> O["### Documentation"]
A --> P["## Release Notes for Previous Versions"]
P --> Q["26.03 → 26.3.0 versioned URL ✓"]
A --> R["## Related Topics"]
R --> S["Helm Charts → pinned 26.05 branch ✓"]
Prompt To Fix All With AI
Fix the following 1 code review issue. Work through them one at a time, proposing concise fixes.
---
### Issue 1 of 1
docs/docs/extraction/releasenotes.md:60-63
Missing section heading for text-to-SQL/tabular content — the bullet on line 63 is orphaned. Every other content category in this release has a `###` subsection heading, and the PR description explicitly lists "text-to-SQL/tabular" as a discoverable highlight subsection. Without the heading, this item falls visually under `### Evaluation` (separated only by a blank line), the page TOC omits it entirely, and readers scanning subsection headers cannot discover the text-to-SQL capabilities.
```suggestion
- BEIR-centric evaluation overhaul and `retriever skill-eval` benchmark CLI for the NeMo Retriever skill (experimental)
### Text-to-SQL and tabular
- Text-to-SQL agent graph and tabular tooling for structured data retrieval, including tabular data ingestion
```
Reviews (8): Last reviewed commit: "Update docs/docs/extraction/releasenotes..." | Re-trigger Greptile
Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com>
…ingest into kheiss/26.05RNs
Expand releasenotes with categorized 26.05 changes, migration notes, RC install guidance, and inline 26.03 section grounded in the 302-commit compare between release branches.
| - Documentation: multimodal extraction is covered on one page with an in-page table of contents and redirects from the former per-topic URLs | ||
| - NeMo Retriever Library currently does not support image captioning via VLM; this feature was added in 26.05 | ||
| - Documentation: multimodal extraction is covered on one page with an in-page table of contents and redirects from the former per-topic URLs | ||
| - Container images built from this repository no longer install `ffmpeg` and `ffprobe` by default. Audio and video extraction require these binaries on `PATH`; for Helm deployments set `service.installFfmpeg=true`, or install system FFmpeg manually in non-container environments. |
There was a problem hiding this comment.
FFmpeg removal attributed to the wrong release
The added bullet — "Container images built from this repository no longer install ffmpeg/ffprobe by default" — describes a 26.05 breaking change (it appears verbatim in the 26.05 "Breaking changes and migration" section on line 29). Placing it in the 26.03 notes implies the removal happened in 26.03. A user running 26.03 who reads this could spend time debugging phantom audio/video failures that won't exist until they upgrade to 26.05, or conclude the 26.03 build is broken when it isn't. Since this is a 26.05 change, remove the bullet from the 26.03 section (the 26.05 breaking-changes section already documents it completely).
Prompt To Fix With AI
This is a comment left during a code review.
Path: docs/docs/extraction/releasenotes.md
Line: 175
Comment:
**FFmpeg removal attributed to the wrong release**
The added bullet — "Container images built from this repository no longer install `ffmpeg`/`ffprobe` by default" — describes a **26.05 breaking change** (it appears verbatim in the 26.05 "Breaking changes and migration" section on line 29). Placing it in the 26.03 notes implies the removal happened in 26.03. A user running 26.03 who reads this could spend time debugging phantom audio/video failures that won't exist until they upgrade to 26.05, or conclude the 26.03 build is broken when it isn't. Since this is a 26.05 change, remove the bullet from the 26.03 section (the 26.05 breaking-changes section already documents it completely).
How can I resolve this? If you propose a fix, please make it concise.Clarify supported vs experimental retriever subcommands per CLI README, and narrow chunking migration to graph/library ingest (service .split() may remain).
Use intro + Helm link + flat Highlights list with nested sub-bullets instead of categorized H3 sections and bold label prefixes.
Resolve releasenotes.md conflict: keep 26.05 release notes and archive 26.03 via Previous Versions link; apply Greptile installFfmpeg YAML fix.
Co-authored-by: Randy Gelhausen <rgelhau@gmail.com>
Co-authored-by: Randy Gelhausen <rgelhau@gmail.com>
Co-authored-by: Randy Gelhausen <rgelhau@gmail.com>
| - BEIR-centric evaluation overhaul and `retriever skill-eval` benchmark CLI for the NeMo Retriever skill (experimental) | ||
|
|
||
|
|
||
| - Text-to-SQL agent graph and tabular tooling for structured data retrieval, including tabular data ingestion |
There was a problem hiding this comment.
Missing section heading for text-to-SQL/tabular content — the bullet on line 63 is orphaned. Every other content category in this release has a
### subsection heading, and the PR description explicitly lists "text-to-SQL/tabular" as a discoverable highlight subsection. Without the heading, this item falls visually under ### Evaluation (separated only by a blank line), the page TOC omits it entirely, and readers scanning subsection headers cannot discover the text-to-SQL capabilities.
| - BEIR-centric evaluation overhaul and `retriever skill-eval` benchmark CLI for the NeMo Retriever skill (experimental) | |
| - Text-to-SQL agent graph and tabular tooling for structured data retrieval, including tabular data ingestion | |
| - BEIR-centric evaluation overhaul and `retriever skill-eval` benchmark CLI for the NeMo Retriever skill (experimental) | |
| ### Text-to-SQL and tabular | |
| - Text-to-SQL agent graph and tabular tooling for structured data retrieval, including tabular data ingestion |
Prompt To Fix With AI
This is a comment left during a code review.
Path: docs/docs/extraction/releasenotes.md
Line: 60-63
Comment:
Missing section heading for text-to-SQL/tabular content — the bullet on line 63 is orphaned. Every other content category in this release has a `###` subsection heading, and the PR description explicitly lists "text-to-SQL/tabular" as a discoverable highlight subsection. Without the heading, this item falls visually under `### Evaluation` (separated only by a blank line), the page TOC omits it entirely, and readers scanning subsection headers cannot discover the text-to-SQL capabilities.
```suggestion
- BEIR-centric evaluation overhaul and `retriever skill-eval` benchmark CLI for the NeMo Retriever skill (experimental)
### Text-to-SQL and tabular
- Text-to-SQL agent graph and tabular tooling for structured data retrieval, including tabular data ingestion
```
How can I resolve this? If you propose a fix, please make it concise.Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
Summary
Adds
docs/docs/extraction/releasenotes.mdas the published 26.05 release notes page for NeMo Retriever Library (26.5.0). The page documents what changed since 26.03 and follows the same presentation pattern as prior extraction release notes (intro, Helm upgrade link, highlights list).26.03 release content is not duplicated on this page; readers are directed to the published 26.03 release notes via Release Notes for Previous Versions.
What's in the doc
.extract(split_config=...), groupedRetriever(...)kwargs (vdb_kwargs,embed_kwargs,rerank), Helmservice.installFfmpeg, Python 3.12Scope
docs/docs/extraction/releasenotes.mdNo product code, Helm values, or workflow changes.
Test plan
releasenotes.mdin the extraction nav without broken internal links (overview.md,deployment-options.md,prerequisites-support-matrix.md)nemo_retrieverAPI (constructor kwargs,split_config, Pythonrequires-python)